Recognition of hand-printed characters using induct machine learning

  • Adnan Amin
  • Aba Rajithan
  • Paul Compton
Learning Methodologies
Part of the Lecture Notes in Computer Science book series (LNCS, volume 1121)


A goal of character recognition is to simplify and automate the development of character recognition algorithms. We describe here an approach based on applying preprocessing to data sets of characters and then applying machine learning to the data sets to build a knowledge base able to classify unseen preprocessed characters. The machine learning method, Induct/RDR, has a number of features which make it particularly suitable for character recognition. It also has the potential to integrate with a manual knowledge acquisition methodology if further refinement is required. Intial results on hand-printed Latin characters show an accuracy of 84% on unseen cases for the machine learning system alone.


  1. 1.
    A. Amin and S. Al-Fedaghi. Machine recognition of printed Arabic text utilising a natural language morphology. Int. Journal of Man-Machine Studies, (6):769–788, 1991.Google Scholar
  2. 2.
    J. Catlett. Ripple down rules as a mediating representation in iteractive induction. In 2nd Japanese knowledge acquisition for knowledge-based system workshop, pages 155–170, 1992.Google Scholar
  3. 3.
    P. Compton and R. Janson. A philosophical basis for knowledge acquisition. Knowledge acquisition, (2):241–257, 1990.Google Scholar
  4. 4.
    G. Edwards et. al. Peirs: a pathologist mainteained expert system for the interpretation of chemical pathology reports. Pathology, (25):27–34, 1993.Google Scholar
  5. 5.
    G. Edwards et. al. An expert system for time course data with expert-managed refinement in context. In 8th Australian Joint Conference on Artificial Intelligence, pages 586–593, 1995.Google Scholar
  6. 6.
    L. Focht and A. Burger. A numeric script recognition processor for postal zip code application. In Int. Conf. Cybernetics and Society, pages 486–492, 1976.Google Scholar
  7. 7.
    H. Freeman. On the encoding of arbitrary geometric configurations. IEEE Trans. Electronic Computer EC-10, (10):260–268, 1968.Google Scholar
  8. 8.
    B. Graines. The trade-off between knowledge and date in knowledge acquisition. In G. Piatetsky-Shapiro and W. Frawley Knowledge Discovery in Databases Cambridge, MA MIT Press, pages 491–505, 1991.Google Scholar
  9. 9.
    B. Graines and P. Compton. Induction of ripple down rules. In 5th Australian Conf. on Artificial Intelligence, pages 349–354, 1992.Google Scholar
  10. 10.
    D. Guillevic and C. Suen. A fast reader scheme. In 2nd Int. Conf. on Document Analysis and Recognition, pages 311–314, 1993.Google Scholar
  11. 11.
    L. Harmon. Automatic recognition of printed and script. Proc. IEEE, (60):1165–1177, 1972.Google Scholar
  12. 12.
    B. K. Jang and R. Janson. One pass parallel thinning: analysis, properties and quantitative evaluation. IEEE Trans. on Pattern Analysis and Machine Intelligence, (11):1129–1140, 1993.Google Scholar
  13. 13.
    R. Plamondon and R. Baron. On-line recognition of handprint schematic pseudocode for automatic fortran code generator. In 8th Int. Conf. on Pattern Recognition, pages 741–745, 1986.Google Scholar
  14. 14.
    J. Quinlan. C4.5: Programs for Machine Learning. San Mateo, CA: Morgan Kauffman, 1993.Google Scholar
  15. 15.
    J. Schuermann. Reading machines. In 6th Int. Conf. on Pattern Recognition, pages 1031–1044, 1982.Google Scholar
  16. 16.
    A. Spanjersberg. Experiments with automatic input of handwritten numerical data into a large administrative system. IEEE Trans. Man and Cybernatics, (4):286–288, 1978.Google Scholar

Copyright information

© Springer-Verlag Berlin Heidelberg 1996

Authors and Affiliations

  • Adnan Amin
    • 1
  • Aba Rajithan
    • 1
  • Paul Compton
    • 1
  1. 1.School of Computer Science and EngineeringUniversity of New South WalesSydneyAustralia

Personalised recommendations